一种整合单通道各向异性扩散信息的水平集图像分割方法*

在对Chan-Vese提出的基于简化Mumford-Shah模型(C-V模型)改进的基础上,针对彩色图像、多光谱图像等多通道图像,提出了一种多通道C-V模型水平集图像分割方法.首先将多通道图像分解到各单通道,使用一种新的各向异性扩散方法对各通道进行平滑滤波,然后使用能够整合各通道各向异性扩散信息的多通道C-V模型进行分割.普通彩色图像与多光谱图像数据的实验结果表明,该方法分割质量明显优于传统的C-V模型分割.

Level set method to image segmentation by integrating anisotropic diffusion information of channels

New level set PDE based on the simplified Mumford-Shah model for image segmentation was proposed by Chan and Vese. A multi-channel C-V model was proposed for multi-channel image segmentation. The method divided the multi-channel image into single channel, and smoothed each channel by an improved anisotropic diffusion method, then used the multi-channel C-V model that could integrate the segmentation information of each channel to segment the image. By practical experiments on color and multispectral images, it is verified that algorithm obtains better results than original method.